CVS head is the right way.
HBMInfoExtractor is very central in this setup so if you look at that and follow who calls it you should be covered.
It is in the tools package, more precisly org.hibernate.mapper.
The Table name completion is something I added very quickly thus the implementation is not the best of breed yet - but look at HibernateNature for the code that fetches the data (I'll be updating this soon as we now have a better way of doing it - but it shows you where the Tables go (and the Tables has columns)
How to build, see
http://www.hibernate.org/268.html
Patches/improvements/bugs should be put in Jira.
don't be afraid to ask ;)
/max